retrofitting of masonry walls by using a mortar joint ...eprints.whiterose.ac.uk/96014/2/revised...
TRANSCRIPT
This is a repository copy of Retrofitting of masonry walls by using a mortar joint technique; experiments and numerical validation.
White Rose Research Online URL for this paper:http://eprints.whiterose.ac.uk/96014/
Version: Accepted Version
Article:
Wang, C, Forth, JP, Nikitas, N orcid.org/0000-0002-6243-052X et al. (1 more author) (2016) Retrofitting of masonry walls by using a mortar joint technique; experiments and numerical validation. Engineering Structures, 117. pp. 58-70. ISSN 0141-0296
https://doi.org/10.1016/j.engstruct.2016.03.001
© 2016, Elsevier. Licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International http://creativecommons.org/licenses/by-nc-nd/4.0/
[email protected]://eprints.whiterose.ac.uk/
Reuse
Items deposited in White Rose Research Online are protected by copyright, with all rights reserved unless indicated otherwise. They may be downloaded and/or printed for private study, or other acts as permitted by national copyright laws. The publisher or other rights holders may allow further reproduction and re-use of the full text version. This is indicated by the licence information on the White Rose Research Online record for the item.
Takedown
If you consider content in White Rose Research Online to be in breach of UK law, please notify us by emailing [email protected] including the URL of the record and the reason for the withdrawal request.
1. Introduction
Masonry is a composite material made of brick units and mortar that has been used for centuries
in building construction. It is in wide usage in seismic-prone areas, especially in the form of infill
panels within reinforced concrete (RC) or steel frames. Therein, infills are customarily considered as
secondary elements (also referred to as non-structural elements) to the structure and are for
simplification not considered in the calculations of seismic capacity. Yet they sustain a large portion of
the energy dissipation [1]. As such, their performance can be a decisive factor leading which may lead
to a catastrophic structural failure. With this in mind, masonry structures often need to be repaired
following earthquake events or enhanced prior to seismic actions in order to ensure that they can
perform their highly sought energy absorption and force relieving roles [2]. In the past decades,
researchers have implemented different methods to enhance the seismic behaviour of unreinforced
masonry walls. These range from the so-called conventional techniques [3] to the latest modern
retrofitting techniques [4].
Conventionally, the surface treatment is an approach to improve the masonry wall behaviour.
Typical surface treatment includes ferrocement, reinforced plaster and shotcrete, with shotcrete being
the most often used method [4]. According to the method, shotcrete overlays are sprayed onto the
surface of a masonry wall over a mesh of reinforcing bars. ElGawady et al. [5] carried out tests on
retrofitted masonry walls by applying shotcrete. The retrofitting was carried out on either one or both
sides, using consistently the same thickness and reinforcement. The test results showed that the
ultimate lateral load resistance of the wall can be increased by a factor of approximately 3. However,
disadvantages of this method include the considerable time required for the implementation and the
adverse impact on the aesthetics of the retrofitted structure.
Grout and epoxy injections are also a broadly used retrofitting approach. The main purpose of
the injections is to restore the original integrity of the retrofitted wall and to fill possible behaviour-
damaging voids and cracks, which are present in the masonry due to physical and chemical
deterioration and/or mechanical actions [3]. The technique was found effective in restoring the initial
stiffness and strength of masonry, while its practicality, relatively minimal cost and easiness of
implementation have rendered it rather popular among engineers. However, any such approach
trivially will be successful only if the mechanical properties together with the physical and chemical
attributes of the employed mix end up being compatible with the masonry to be retrofitted [6].
Some of the drawbacks of the quoted conventional methods can be overcome by the Fibre
Reinforced Polymer (FRP) reinforcement. Retrofitting of unreinforced masonry walls using FRP can
increase the lateral resistance by a factor ranging from 1.1 to over 3 [4]. Alcainoand Santa-Maria [7]
presented experimental results from clay brick masonry walls retrofitted with carbon fibre. From the
results analysis, it was found that the strength of the walls could increase between 13% and 84%.
Also, Mohmoodand Ingham [8] conducted a research program in order to investigate the
effectiveness of FRP additions as seismic retrofit interventions for in-plane loaded unreinforced
masonry walls. The experimental results showed that the shear strength increased up to a factor of
3.25. In general, the retrofitting of masonry walls using FRP material addition has the common
advantage of little added mass while mostly producing low disturbance for achieving a relatively high
improvement in strength. However, the main drawbacks are the high cost, the high technical skill
required for their installation, the affecting of architectural aesthetics and the not so broad experience
with these materials particularly relevant to their aging.
To the authors’ knowledge, most of the research on the mechanical behaviour of masonry and
the retrofitting measures were focused on single-leaf walls, with only very few exemptions expanding
on double-leaf or multi-leaf masonry walls. Predicting the behaviour of multiple-leaf masonry walls is a
challenging issue, given the influence of a wide range of factors as the mechanical properties of the
leaves, their dimensions and the way they are connected to each other. Still, double-leaf walls can be
found in many historic structures as well as in modern structures and they have regularly been
exposed to considerable earthquakes obviously affecting the holistic structural dynamic performance.
Therefore, it feels necessary to also conduct research on such a construction system shedding light to
previous gaps in knowledge. Anand and Yalamanchili [9] analysed a composite masonry wall made of
block and brick units and tied together by two different in thickness collar joints, 9.5 mm and 51 mm.
The composite masonry walls were subjected to both vertical and horizontal loads in a 3D
arrangement. From the results analysis, it was found that the collar joint failure is brittle in nature.
Pian-Henriques et al. [10] conducted a series of experimental tests on multi-leaf masonry wall panels
under a combined shear and compression load with the aim to predict their load carrying capacity and
failure mode. The specimens consisted of two external leaves made of stone blocks bonded together
with mortar joints while the internal leaf consisted of a mixture of mortar with stone aggregates. A
simplified calculation for predicting the compressive strength of composite walls has been presented
good agreement with experimental results obtained. Ramalho et al. [11] modelled the experimental
specimens of Pian-Henriques et al. [10] by applying a damage model which was developed to
interpret the time evolution of mechanical damage in brittle materials. The models were implemented
using the finite element codes ABAQUS and FEAP and comparisons made on the results obtained.
The proposed numerical codes were able to capture the different features of the nonlinear mechanical
behaviour of multi-leaf walls. However, as perfect bonding was assumed between the adjacent layers
during the modelling, some of the numerical results were overestimated. Also, Binda et al. [12]
conducted research on multi-leaf masonry walls in order to understand the load-transfer mechanisms
between the individual walls although the collar joint which was used for the construction of the walls
were much thicker than what is suggested in British Standard 5628-1:2005 [13] (i.e. the space
between two parallel single-leaf walls is to not exceed 25mm).
In this paper, a conventional though practical, novel retrofitting approach is introduced. Namely,
the traditional method of building a wall parallel to an existing single-leaf wall and bonding the two
together using a mortar collar joint is being considered as a possible strengthening and retrofitting
technique. The method does not require sophisticated workmanship because of its easy
implementation, which further renders it cost-effective.
In general, the application can be divided into two categories: a) the pre-damage enhancement;
and b) the post-damage repairing. Earthquake being a specific very interesting catastrophic damage
case with great relevance to masonry wall failures is what will be particularly discussed hereafter. For
the purpose of the specific project in pre-earthquake enhancement tested walls, the second wall was
built parallel to the existing one and bonded with a relatively thin collar joint before the test. For the
case of post-earthquake repaired walls, the second wall was attached to the existing one after it had
been tested (and as such partially damaged). The collar joint dimensions were kept constant while the
damage progressed only to the very early plastic range (i.e. cracking far from collapse). A preliminary
parametric study has been conducted to evaluate the performance of the enhancement method using
a monotonically increasing quasi-static loading scheme. Notably, the whole study is not only relevant
to earthquake engineering, which is a rarity in UK; double-leaf (collar jointed) walls can also be used
to improve a structure’s lateral stability (e.g. against wind or blast loading) through adding stiffness
[14]. Thus, this research broadly aims to generate knowledge and understanding which can be
directly applied in a number of structural applications.
On the numerical modelling side, in the past decades, research relevant to masonry walls has
been advanced considerably. However, the modelling of a load bearing masonry wall or masonry infill
under in-plane combined loading remains difficult primarily due to the complex mechanics developed
within the different materials of the wall. A number of different approaches have been implemented to
simulate the mechanical behaviour of masonry walls subjected to static or dynamic loading that can
act in-plane, out-of-plane or even simultaneously in both planes. The selection of the most
appropriate method to use depends on, among other factors, on the structure under analysis; the
level of accuracy and simplicity desired; the knowledge of the input properties in the model and the
experimental data available; the amount of financial resources; time requirements and the experience
of the modeller (Lourenco [15]). Preferably, the approach selected to model masonry should provide
the desired information in a reliable manner within an acceptable degree of accuracy and with least
cost. According to Lourenco [15], the available strategies for the numerical modelling of masonry
structures would fall within one of two categories: a) micro-scale; and b) macro-scale modelling.
In macro-scale modelling, the masonry units and mortar joints are smeared into an averaged
continuum. There are no distinctions between the units, the mortar and their interfaces. This model
can be applicable when the dimensions of a structure are large enough, compared to the constituent
parts, so that a description involving average stresses and strains becomes acceptable [16].
Considerable computational time can be saved by applying this method. However, unconditionally
accurate results and fine-detail of the behaviour cannot be captured by the nature of this approach.
On the other hand, the micro-scale modelling can be split into the following two approaches: a)
simplified micro-modelling; and detailed micro-modelling. In the simplified micro-modelling approach
expanded units are modelled as continuous elements while the behaviour of the mortar joints and
unit-mortar interface is lumped in discontinuous elements. In the detailed micro-modelling approach
both the masonry units and the mortar are discretised and modelled with continuum elements while
the unit/mortar interface is represented by discontinuous elements accounting for potential crack or
slip planes. Detailed micro-modelling is probably the most accurate tool available today to simulate
the real behaviour of masonry as the elastic and inelastic properties of both the units and the mortar
can be realistically taken into account. With this method, a suitable constitutive law is introduced in
order to reproduce not only the behaviour of the masonry units and mortar, but also their interaction.
However, any analysis with this level of refinement requires large computational effort to analyse.
Thus, this method is used mainly to simulate tests on small specimens in order to determine
accurately the stress distribution in the masonry materials. The drawback of the large computational
effort required by detailed micro-modelling is partially overcome by the simplified micro-modelling
strategy. In this case, each joint, consisting of mortar and the two unit-mortar interfaces, is lumped
into an “average” interface while the units are slightly expanded in size in order to keep the geometry
unchanged. Within this approach, it is possible to consider masonry as a set of elastic blocks bonded
together by potential fracture slip lines at the joints. The main methods available for modelling
masonry structures using the simplified micro-modelling approach include: a) the discontinuous finite
element method; and b) the discrete element method.
When modelling masonry using the discontinuum finite element method, discontinuities are
generally introduced using interface elements, for which the constitutive model is in direct relation with
the stress vector and the relative displacement vector along the interface (Oliveira [33]). Thus, for an
accurate simulation of masonry behaviour, it is essential to obtain a constitutive model for the
interface elements which is able to capture realistically the behaviour of masonry and be able to
simulate all the failure mechanisms. Page [17] first introduced masonry as a two-phase material,
which translates to the bricks taken as linear elastic and the mortar-brick interface taken as inelastic
obeying to a simple Mohr-Coulomb failure criterion. Lourenco [18] subsequently introduced a
compressive cap to the failure surface in Page’s model. By this crushing of the masonry bricks is also
enabled beyond the interfaces, allowing for all possible failure models to be taken into account.
Although a micro-scale model needs more computational time, it can let many salient behaviour
features to emerge, thus giving a better understanding and predicting insight of the masonry walls’
performance. Al-Chaar and Mehrabi [19] modelled a few RC frames infilled with masonry walls using
this method in DIANA. Furthermore, a lot more researchers have applied this method to model
masonry structures. (Van Zijl [20], Dolatshahi and Aref [21])
The discrete element method (DEM) is characterized by modelling the materials as an
assemblage of distinct blocks or particles interacting along their boundaries. The formulation of the
method was proposed initially by Cundall [22] for the study of jointed rock, modelled as an
assemblage of rigid blocks. Later this approach was extended to other fields of engineering requiring
a detailed study of the contact between blocks or particles such as soil and other granular materials
(Ghaboussi and Barbosa [32]. In the last two decades, the approach was applied successfully to
model masonry structures by Lemos [23] and Giamundo et al. [26] in which the collapse modes were
typically governed by mechanisms in which the deformability of the blocks plays little or no role. Also,
the possibility of frequent changes in the connectivity and the type of contact as well as marked non-
linearity induced by the inability of the masonry joints to withstand tension makes DE a suitable
method for solving problems involving discontinuities as is the case with low bond strength masonry
(Sarhosis & Sheng [24] and Sarhosis et al. [25]).
However, nowadays, in modern FE, the time integration algorithm might be explicit or implicit,
the contact size and extent is updated, large or small displacements and rotations can be taken into
account, the contact detection algorithms detect new contacts and even self-contact, the contact
algorithms are much more sophisticated and accurate than the classical DE contact strategies etc.
Therefore, the use of the computational strategy to use is rather a matter of taste provided that the
user is experienced.
The development of a computational model based on the discontinuum finite element approach
is presented here. Namely the bricks were modelled as rigid elements separated by zero thickness
interfaces representing the mortar joints. The interface inelastic properties were simulated using a
Mohr-Coulomb failure surface combined with a tension cut-off and a compression cap. The modelling
approach when referring to a single-leaf masonry wall panel (i.e. unretrofitted) focused only in two
dimensional analyses while for double-leaf masonry walls it is clear that it was necessary to expand in
three dimensional one. The model was implemented in the commercial software MIDAS FEA [26] and
all analytical results were verified and validated against currently derived experimental outputs.
2. Experimental work
A series of single and double leaf brickwork masonry wall panels were tested in the laboratory. The
experimental campaign included four tests on single-leaf and three on double-leaf walls. The
experimental observations were primarily focused on static displacement and load capacities clearly
supporting a quasi-static rationale for performing any earthquake load related assessments. The in-
plane dimensions of each brickwork masonry wall panel tested in the laboratory were 975 mm x 900
mm. All panels were built with stretcher bonded brickwork and rested on a steel base-plate which was
constrained by a steel portal.
2.1. Materials
2.1.1. Brick
All the test panels were constructed with UK standard size 215 mm x 102.5 mm x 65 mm
Engineering Class B perforated bricks. From the manufacturer’s specification, the bricks had average
water absorption of 5.6% ( 0.6%), porosity equal to 25%, density of approximately 1885kg/m3 and
compressive strength of the order of 35 N/mm2. A series of small scale tests to investigate the
modulus of elasticity of bricks have been performed according to BS 3921:1985 [27] and BS EN772-
1:2011[28]. In total, ten samples were tested and the elastic modulus of the bricks was found to be
equal to 19.9 kN/mm2.
2.1.2. Mortar
There were two different types of mortar used in the experiments. These are: a) Type S; and b)
Type N. Type S had mix proportions of Ordinary Portland Cement (OPC) : lime : sand equal to
1:1/2:4½. The compressive strength for 100 mm cubes cured in a fog room with 99% relative humidity
and 21 °C temperature was 12.7 MPa ( 1.2MPa) (based on eight tests). Type N had mix proportions
of Ordinary Portland Cement (OPC) : lime : sand equal to 1:1:6. The compressive strength for 100
mm cubes cured in a fog room with relative humidity 99% and temperature 21 °C was 6.7 MPa ( 0.4
MPa) (based on eight tests). All mortar compressive strength tests were undertaken as per BS EN
1015-11:1999 [29]. The mortar joints were all nominally 10 mm thick.
2.2. Panels’ tests arrangements and procedure
Each wall panel was tested by applying an external load on the top-left hand corner. The load
was applied to each panel using a hydraulic ram and was distributed through a thick steel spreader
plate which was embedded in mortar on the surface of the brickwork. The steel plates were spanning
over the top three courses in the vertical direction and over one brick length horizontally. There was a
10 mm gap initially, which then increased to 20mm, between the unloaded side of the panel and the
portal frame column in order to provide clearance for horizontal displacements. For the first three
courses, starting from the base, this gap was filled with mortar to restrict any horizontal movement
allowance of the wall.
A horizontal load was applied to each wall incrementally (2 kN/min) until the panel could no
longer carry the applied load. Among others, the scope of the test rig was to potentially simulate the
RC frame restraint as experienced by a real infill wall. Therefore, a vertical load cell was also used to
suppress the vertical uplift of the restrained leaf, mimicking the interaction with an RC frame. Initially
the vertical load was set to 20 kN (and subsequently adequately increased with target to limit any
base rotation). Furthermore, a LVDT was set up on the top hand corner of the panel. At each load
increment, the LVDT’s readings were monitored for signs of continued gradual increase in deflection
under constant load. This re-distribution of stress was particularly noticeable as cracks developed and
propagated through each panel. During the test, surface cracking was inspected visually. Typically the
first visible cracks were of the order of 0.2mm wide. The test rig for the single-leaf wall is shown in
Figure 1.
A second series of tests was subsequently carried out for all double-leaf walls on an updated
apparatus (Figure 2). The second leaf was built parallel to the existing one and was ‘tied’ to it using a
10mm thick collar joint. The material of the collar joint was exactly the same to the mortar joint used
for the construction of the second leaf wall. Mortar was successively filling up to the bricks’ top and
the collar joint after constructing each new layer of bricks. Therefore, it can be simply assumed that
any holes in the bricks and collar joint between the two walls were fully filled with mortar. There was
no surface treatment applied on the walls. Also, the new panel was not restricted in any way by the
portal frame, which meant that it could move freely throughout its length along its in-plane axis. The
load was solely applied to the initial panel which was restricted by the portal frame, while the loading
setup was exactly the same to the single-leaf case. Thus, there was no direct loading applied to the
second leaf wall; the only load sustained was transferred by some shear transfer mechanism from the
initial panel.
For the double-leaf walls, there was a further division into two categories relevant to their
damage stage. These will be quoted as “pre-earthquake” and “post-earthquake”. In the pre-
earthquake case, the second-leaf wall was attached to the first leaf wall before it even got any load.
Ideally, any practical retrofit action is performed much later than the initial structural installation. Still at
this level a sensible problem reduction approach needs to first address the concurrent wall installation
as this being the simplest possible answer to the complexity of the aging process. As a matter of fact
this rationale excludes the parameter of differential aging from the multivariate problem in-hand. In the
post-earthquake case, the second leaf was attached to the first one only after the latter had nominally
failed making it essentially a means of repair. In all cases, walls had been cured for 14 days under a
polythene sheet before being loaded. To this testing rule there was only one exception. The referred
to as wall W6 (a single-leaf wall) was cured for an extended period of 42 days before being tested in
order to have some indication of the curing impact and a baseline for the efficacy of any later
remediation. The preparatory single-leaf test was interrupted when initial fine cracking occurred.
Subsequently it did not get any crack repair, since these were of hairline nature, and got retrofitted by
“attaching” a second wall to it using the previously discussed mortar collar joint technique (becoming
then Wall 7). A summary of the full test configurations studied, indicating the adopted tests’ naming
conventions for any later reference is provided in Table 1.
Figure 1. Testing rig of single-leaf panel
Figure 2. Testing rig of double-leaf panel
Table 1. Summary of test results
Wall name Wall type Mortar type Days cured Case
W1 Single-leaf S 14 -
W2 Single-leaf S 14 -
W3 Single-leaf N 14 -
W4 Double-leaf N 14 Pre-earthquake
W5 Double-leaf N 14 Pre-earthquake
W6 Single-leaf N 42 -
W7 Double-leaf N 14 Post-earthquake
3. Experimental results
3.1. Failure patterns; an initial qualitative assessment
3.1.1. Single-leaf walls
(a) Test panel W1 (b) Test panel W3
Figure 3. Failure pattern of single-leaf walls
Figure 3 shows the failure pattern of the single-leaf masonry wall panels W1 and W3. Very
similar failure patterns were observed for the other single-leaf masonry walls (W2 and W6). According
to the failure patterns shown in Figure 3, the failure mode of a single-leaf wall can be described by a
major diagonal crack. Before this diagonal crack developed, some small, hairline shear cracks
appeared along the bed joint length. Further, with increasing horizontal load, the top corner of the wall
(indicated as area 1 in Figure 3) began to crush and cracks started propagating from that region down
to the base of the wall. Stresses kept increasing with applied load and once the strength of the
masonry was exceeded the failure occurred in the form of the earlier quoted diagonal crack spanning
widely from area 1 to area 3, following a staircase path along the mortar interfaces. This typical
mechanical behaviour of a masonry wall under lateral load can also be seen in the work of
Vermeltfoort et al. [30]. The wall point at the top of the edge gap-filling mortar in area 3 is clearly a
point of rotation and as expected no local extensive crushing of the masonry was observed below this
region. In this experimental study, the occurrence of the diagonal crack signified the end of the testing.
However, in practice a masonry panel loaded in-plane within a frame will become locked in and
continue resisting the panel deformation even after the diagonal cracks are formed; the most notable
aspect of such a scenario is the potential for additional energy dissipation (Mehrabi et al. [1]) allowed
within the restrained sliding of the damaged interfaces. These tests do not consider any load cycling
or dynamic effect that is critical for assessing holistically along these lines the masonry performance.
Still they constitute an insightful first attempt to explain and comprehend the up-to failure performance
of the masonry wall.
3.1.2. Double-leaf walls
(a) Front side (loaded wall) (b) Back side (parallel wall)
Figure 4. Failure pattern of double-leaf wall W5
(a) Front side (loaded wall) (b) Back side (parallel wall)
Figure 5. Failure pattern of the double-leaf wall W7
3.1.2a Pre-earthquake test
Figure 4 shows the failure mode of the wall panel W5. From Figure 4, it is clear that the pre-
earthquake double-leaf walls failed by diagonal cracking, similar to the single-leaf wall case. However,
at this instance walls had more cracks than their single-leaf counterpart prior to the formation of the
decisive diagonal crack that signified the ultimate failure; this is a sign that for the double-leaf walls,
ductility (i.e. extend of plastic deformation) had improved through the presence of a second leaf. In
terms of the failure process, first, some small hairline cracks appeared along the bed joints on both
leaves; similar to the single-leaf wall previously. Note that the cracks in the second leaf appeared later
than the ones on the first leaf. Further, in all cases and at all times the cracks of the second leaf were
less compared to the cracks of the first leaf. Thus, it becomes apparent that the stress transfer
between the two leaves was effective throughout the different loading stages as initially envisaged.
Namely, the load is applied directly to the first wall and distributed to the second wall consistently via
the collar-joint. Therefore, although the two leaves are joined and the width of the loaded area is
effectively close to double the initial, the real stress is not distributed evenly, being concentrated at the
top corner of the first wall and “flowing” inhomogeneously into the second wall. The uneven
distribution of the stresses between the two walls is also influenced by the boundary conditions
imposed. The second leaf is not restrained by the gap-filling mortar and therefore is becoming less
stiff. The two walls are bonded together acting in a way as a composite construction (Figure 6a).
3.1.2b Post-earthquake test
Figure 5 shows the failure mode as observed in the double leaf brickwork masonry wall panel
W7. It can be seen that the first leaf of the post-damaged masonry wall panel behaved in a similar
manner to the single-leaf masonry wall panels tested previously were failure was governed by a wide
diagonal crack. This is obviously affected strongly by the preloading and incipient damage induced to
the wall. However, the second wall behaved quite differently to what was seen before. The actual
failure for this case was established by a horizontal shear crack, initiated by the failure of the collar
joint. The collar joint actually detached itself from the first wall whilst remaining connected to the
second wall (Figure 6b). Therefore, the collar joint did not manage to sustain the integrity of the
construction throughout the experiment. The composite masonry wall constituents nearly work
individually after the first wall debonded from the second. The cracking pattern succession observed
in the first wall was also very different to that seen in the second wall. In the front side of the wall
panel W7 diagonal cracks passed through the mortar joints and crossed the bricks. However, in the
back side of the wall panel W7, only a small sliding and stepped crack appeared at the bottom of the
wall. The localization of this sliding and stepped crack must intuitively follow a weakest link path
through the mortar joints.
(a) Top side of W5 (b) Top side of W7
Figure 6. Failure patter of the collar joint
3.2. Failure load and deflection
Table 2. Failure load and deflection of all tests
Test No. Wall type
Horizontal load (kN)
Displacement at yield point (mm)
Maximum displacement
(mm)
Mortar compressive strength (MPa)
Case
W1 Single-leaf 58 9.7 13.1 12.7 -
W2 Single-leaf 64 10.1 11.2 15.3 -
W3 Single-leaf 70 8.2 20.0 6.7 -
W4 Double-leaf 91 10.1 11.4 6.3 Pre-earthquake
W5 Double-leaf 93 10.3 12.6 6.6 Pre-earthquake
W6 Single-leaf 75 9.03 9.03 8.1 -
W7 Double-leaf 77 8.8 17.6 7.1 Post-earthquake
Figure 7. Load-Deflection relationship of single-leaf walls
The ultimate failure loads along with critical deflection parameters for all tests are summarised in
Table 2. The load against deflection curves for the ensemble of single-leaf walls is shown in Figure 7.
The stiffness of wall W1 is very similar to, although slightly lower than that of W2. More importantly
some extensive capability for plastic deformation is observed in wall W1 while this was not the case
for W2. W1 could deform even more and its full plastic range was not pursued since the limitation of
the apparatus clearance was reached (this was increased only thereafter). Such experimental
deviations are expected in similar masonry constructions, yet this seems to be quite a substantial
difference that should probably get attributed to a substantial material deviation that wasn’t identify.
When referring to the different mortar type wall W3 (i.e. Type N, see Table 1) all strength and
deformation capacity variables increased consistently and substantially.
The testing of wall W6 was stopped when it was nominally assumed to have yielded; this state
was taken at the point when initial ‘fine’ cracking appeared and the horizontal load-deflection relation
started deviating increasingly from the initial elastic region. At that point W6 was unloaded and its
damaged stage was considered the base for the later post-damage retrofitting study. Figure 7 reveals
that there was no considerable post-peak load behaviour captured for W6 as intended (see elastic
recovery). The stiffness of wall W6 was evidently greater than that of W1 and W2. Although this can
be attributed to the increased curing time (i.e. compared to W1 and W2, W6 was cured for 42 days
instead of 14 days) this increased stiffness being also apparent in the case of W3 seems mainly a
product of the different mortar type. However, the stiffness of the W6 is lower than that of W3, further
implying the small effect of additional curing times beyond a certain duration.
Figure 8. Load-Deflection relationship the of double-leaf walls
Figure 8 illustrates the horizontal load-deflection behaviour for all the collar-jointed masonry walls.
Walls W4 and W5 (pre-earthquake method) exhibited a much higher failure load (91 kN and 93 kN,
respectively) than any of the single leaf walls, which failed at loads ranging between 58 kN to 70 kN.
Although W4 and W5 have similar failure loads, yet their ultimate deflection differs. This is an artificial
output with the measurement of W4 encompassing a slippage contribution without which the
displacement behaviour becomes quite alike to W5 rendering any remaining difference falling within
the acceptable experimental deviation bands. Interestingly, wall W7 (the post-earthquake retrofit wall),
although only achieving a failure load more in-line with the single-leaf walls (around 80 kN) going
approximately only halfway through the capability gains of the pre-earthquake retrofit method, exhibits
sustained ductility with much more gradual/reduced softening.
Figure 9. Load-Deflection relationship of post-earthquake strengthening
Figure 9 presents the load against deflection curves for walls W6 and W7. It can be seen that
although the failure load of the repaired and strengthened double-leaf wall (W7) was not substantially
increased, the initial stiffness has been improved significantly; it actually reached a value almost twice
the single-leaf one (W6). For the repaired double-leaf wall (W7), a significant amount of ductility was
observed, as previously quoted, yet, this is not strictly comparable to the single leaf damaged variant.
This is due to the fact that testing of W6 stopped when only initial cracks appeared on its body
naturally yielding that the amount of ductility represented in Figure 9 is not indicative of its full capacity.
4. Development of the computational models for masonry
The development of computational models to simulate the mechanical behaviour of the brickwork
masonry wall panels tested in the laboratory is presented hereafter. The devised models were based
on the micro-scale modelling approach (Lourenco [15]) and were developed using the commercial
software MIDAS FEA [18]. For the development of the numerical model, the mortar joints were
lumped into a zero-thickness interface while the dimensions of the brick units were expanded to keep
the geometry of the given masonry structure unchanged. In addition, vertical predefined cracks were
assigned to the middle of each brick element (Figure 10). This is due to the fact that in masonry
structures, as also evidenced in the current experimental failure patterns, most of the propagating
cracks beyond being located in the mortar they also develop across the bricks [24, 25].
For the 2D models, practiced in the case of single leaf walls, the brick units were represented
by eight-noded plane stress continuum elements, while the brick-mortar interfaces were represented
by six-noded line interface elements. For the 3D models relevant to collar jointed walls, the bricks
were represented by eight-node hexahedron solid elements while surface interface elements were
used to analyse the interface behaviour between the solid brick elements. Bricks were assumed to
behave in a purely elastic manner while the joint interfaces were simulated using a Mohr-Coulomb
failure surface combined with a tension cut-off and a compression cap. For all the cases, the size of
the mesh was kept at less than 50 mm.
Figure 10. Micro-modelling strategy for masonry (after Lourenco [15])
4.1 Single-leaf walls; comparison to experiments
For the 2D geometrical models representing the single-leaf brickwork wall panels tested in the
laboratory the consideration of the 10mm thick mortar joints was taken by increasing by 5mm in each
Potential brick crack Brick-mortar interface
relevant face direction the brick size in order to give a typical MIDAS block size of 225mm x 102.5mm
x 75mm (see Figures 10 and 11 to identify the model geometry).
Figure 11. The validation 2D model geometry in MIDAS FEA
4.1.1 Material parameters
For brick-mortar interfaces, the normal stiffness ( nk ) and shear stiffness ( sk ) are very difficult to
obtain. Therefore, they are both assumed based on previous literature values (see Lourenco [15],
[18]). An extensive study on the mechanical behaviour of brick-mortar has been conducted by Van der
Pluijm [31]. Van der Pluijm found that the bond strength tf varies between 0.3 to 0.9 N/mm2 and the
mode I fracture energy IfG , from 0.005 to 0.02 Nmm/mm2. BS 5628: 2005 [13] gives design values
for cohesion C ranging from 0.35 to 1.75 N/mm2 and tan for friction equal to 0.6. Van der Pluijm
found that the Mode II fracture energy IIfG ranges from 0.01 to 0.25 Nmm/mm2 while the initial
cohesion value ranges from 0.1 to 1.8 N/mm2. If only one of the fracture mode energy is given, then
the other one can be obtained simply as 10II If fG G . In addition, Van der Pluijm found that the
tangent of the initial internal friction angle 0tan ranges from 0.7 to 1.2 for different unit/mortar
combinations. The tangent of the residual internal friction angle tan r was approximately constant
and equal to 0.75. The average value for dilatancy angle tan ranges from 0.2 to 0.7 depending on
the roughness of the brick surface for low confining pressure. Based on the above background and
some preliminary testing attempts all the material values required for the implementation of the single-
leaf wall in MIDAS FEA are shown in Tables 3 and 4.
Table 3. Properties of the brick masonry units.
Variable Value
Unit Weight d [kg/m3] 1885
Young Modulus E [kN/mm2] 19.9
Poisson’s Ratio v 0.15
Table 4. Properties of interfaces
Variable Brick-mortar interface
Brick interface representing potential vertical crack
Normal stiffness nk ( N/mm3) 13 1000
Normal stiffness sk ( N/mm3) 5.3 1000
Tensile strength tf ( N/mm2) 0.4 2
Tensile fracture energy IfG ( Nmm/mm2)
0.022 0.08
Cohesion C ( N/mm2) 0.56 -
Friction coefficient tan 0.75 -
Dilatancy coefficient tan 0.56 -
Shear fracture energy IIfG ( Nmm/mm2)
0.175 -
Compressive strength cf ( N/mm2) 8.5 -
Compressive fracture energy cG (Nmm/mm2) 5.0 -
Compressive plastic strain at cf 0.093 -
4.1.2 Modelling results
Figure 11 shows a qualitative comparison of the failure patterns developed in the numerical and
experimental tests. The relevant horizontal load-deflection curves for the wall W3 are presented in
Figure 12. In Figure 12 it can be seen that the computational model is capable to capture with
sufficient accuracy the failure mode. Further, Figure 13 compares the experimental against the
numerical load against displacement curves. In Figure 13, the computational model is able to capture
the stiffness of the wall obtained experimentally for the elastic and incipient plastic deformation
regions. This translates to a load reaching above 55kN. At that point, some deviation starts
developing with the numerical model showing some first cracking signs which still do not affect
majorly the stiffness which continues unreduced towards reaching a maximum load value in close
proximity to the experimental observations. The wall showed a distinctively different behaviour
regarding the plastic region with the numerical analogue showing its collapse soon after reaching the
ultimate load whereas the experimental case presented increased ductility. Very interestingly shape-
wise this looks like the difference between wall panels W1 and W2 observed experimentally earlier.
Such a difference yet may well be caused by the difference with which the frame restraint was
numerically realised. Relevant to this probably one should note the correct cracking patterns that
extend accurately beyond the mortar through bricks in all cases.
(a) Analytical result (b) Experimental result
Figure 12. Failure pattern of single-leaf masonry wall W3
Figure 13. Load-deflection of wall W3
4.2 Double-leaf walls (pre-earthquake); comparison to experiments
3D geometrical models representing the double leaf brickwork panels connected with collar joint
were created again in MIDAS FEA (Figure 14) using similar to the 2D previous assumptions (e.g. a
potential crack was again placed in the middle part of the bricks). Trivially the behaviour of the collar
joint is decisive for the overall behaviour of the wall. Likewise the mortar joint, the collar joint was
smeared into an interfacial element of zero thickness.
4.2.1 Material parameters
Since the two merged walls have been constructed at the same time and were cured under
identical conditions, their material properties would be the same. In this circumstance, the behaviour
of the interfaces pertaining to the two leaves can be modelled with the same interface property as the
single-leaf wall shown in the above section (see Table 3). However, the material property of the collar
joint is not known apriori. As the mortar used in the collar joint is the same to that used on the bed and
head joints, the property can be assumed similarly. However, unlike bed/head joints, the collar joint
was not compressed by the external vertical or horizontal load, resulting in the normal/shear stiffness
and cohesion to be relatively smaller. The additional material parameters artificially reduced to comply
with this rationale are presented in Table 5.
Figure 14. The validation 3D model in MIDAS FEA
Table 5. Properties of different interfaces
Variable Collar joint
Normal stiffness nk ( N/mm3) 9
Shear stiffness sk ( N/mm3) 3.6
Tensile strength tf ( N/mm2) 0.22
Tensile fracture energy IfG ( Nmm/mm2)
0.018
Cohesion C ( N/mm2) 0.30
Friction coefficient tan 0.75
Dilatancy coefficient tan 0.56
Shear fracture energy IIfG ( Nmm/mm2)
0.17
Compressive strength cf ( N/mm2) 8.5
Compressive fracture energy cG ( Nmm/mm2) 5.0
Compressive plastic strain at cf 0.093
4.2.2 Modelling results
Figures 15 and 16 show the qualitative comparison of failure patterns developed in both the
numerical and experimental scenarios. Macroscopically the cracks seem to have occurred in similar
positions and formations on both sides. Still there is some difference when one observes in detail the
developed cracks (extend of cracks, passing through bricks etc.). Such differences most probably
originate due to local weak zones in the built wall. Still the match is acceptable indicating the good
approximation in modelling the collar joint and the inherent variability in masonry materials. In Figure
17, the relevant load against deflection curves for wall panel W4, in both the analytical and
experimental cases, showed really close behaviour concerning all the critical parameters (i.e.
maximum load, failure deflection and stiffness). In the numerical case, when comparing to the single
leaf wall studied before the developed plasticity substantially increased, yet it still couldn’t reach to the
levels observed during the experiment.
(a) Experimental (b) Analytical
Figure 15. Failure pattern of double-leaf masonry wall 4 on the back side
(a) Experimental (b) Analytical
Figure 16. Failure pattern of double-leaf masonry wall 4 on the front side
Figure 17. Load-deflection curve of Wall 4
4.3 Double-leaf walls (post-earthquake); comparison to experiments
For the masonry wall W7 the damage introduction results to some interesting modelling
idiosyncrasies. The existence of some initial minute cracks in the first wall need to also be
approximated correctly if an accurate behaviour is to surface from the modelling attempt. Based on
the experimental observations, a grid of existing cracks was pre-defined. This is represented by
dashed lines in Figure 18, showing the geometry of the numerical implementation of the wall. In this
model, the cracks were assumed to not have any interaction (though there might by some residual
friction within them). The boundary conditions were envisaged to be identical to the previous double-
leaf wall setup (i.e. pre-earthquake).
Figure 18. (a)Cracks in experimental results (b)pre-defined cracks in finite element modelling
4.2.3 Materials’ parameters
In this case with the first wall having been cured for 42 days and the second having been cured
for only 14 days, the two leaves had to be modelled differently. Thus, the brick-mortar interface for the
two walls acquired different parameters. To this one should further note the need for different
modelling of the collar-joint that was of particular significance in this study. The first leaf owing to its
additional curing time when compared to the second leaf was considered having more strength. The
joint properties within the second leaf were taken identical to the previous single-leaf walls. Contrary
to the previous approach where the collar joint has been uniformly smeared out now it wasn’t due to
the fact that the interface between the first leaf and the collar joint (interface 1) is different to the one
between the second leaf and the collar joint (interface 2) (see Figure 19). Namely the collar joint being
built at the same time with the second leaf, was considered with properties identical to the pre-
earthquake collar joint case, while the connection with the first leaf was considered minutely weaker.
Based on these assumptions, the now extended list of material parameters are given in Table 6.
Table 6. Properties of the zero thick ness interfaces (to read in conjunction with Figure 19)
Variable 1st leaf
mortar
2nd leaf
mortar
Collar joint
interface 1
Collar joint
interface 2
Normal stiffness nk ( N/mm3) 15 13 8.5 9
Shear stiffness sk ( N/mm3) 6.0 5.3 3.5 3.6
Tensile strength tf ( N/mm2) 0.45 0.4 0.22 0.22
Tensile fracture energy IfG ( Nmm/mm2)
0.022 0.022 0.018 0.018
Cohesion C ( N/mm2) 0.56 0.56 0.3 0.3
Friction coefficient tan 0.75 0.75 0.75 0.75
Dilatancy coefficient tan 0.56 0.56 0.56 0.56
Shear fracture energy IIfG ( Nmm/mm2)
0.175 0.175 0.2 0.17
Compressive strength cf ( N/mm2) 8.5 8.5 8.5 8.5
Compressive fracture energy cG ( Nmm/mm2) 5.0 5.0 5.0 5
Compressive plastic strain at cf 0.093 0.093 0.093 0.093
Figure 19. Plan view of the double leaf brickwork masonry wall illustrating the collar joint interface 1 and 2 (To read in
conjunction with Figure 2).
4.2.4 Modelling results
The failure patterns for the retrofitted wall in all cases are shown in Figures 20 and 21. Again
macroscopically the match is quite accurate between model and numerical prediction, particularly
when considering the somewhat approximative character of the interface properties adopted. The
overall extend of the cracks and their distributions in the different leaves are in good agreement while
salient features such as the exact crack position and length differ. The newly built wall side, although
being considered perfectly bonded to the initial side it could not actually relieve it considerably from
stresses. Without carrying much load, as evidenced from its minor damage, it followed the whole
wall’s failure being ultimately quite intact with minimal cracking. As the front side was pre-damaged
and also carried most of the loading during the test in the end it got completely damaged, by forming
two main step-like diagonal cracks. This failure mode was well captured numerically. Furthermore,
there appeared to be more cracks compared with the pre-earthquake scenario. Figure 22 shows the
horizontal load-deflection relationship of W7. Therein, the analytical modelling shows an extensive
drop in load after yielding and a much reduced plastic zone as also seen earlier. Still the ultimate load
and stiffness, which is substantially improved compared to the single leaf case, are well predicted.
(a) Experimental (b) Analytical
Figure 20. Failure pattern of double-leaf masonry wall on the back side of W7
Collar joint interface 1
Collar joint interface 2
(a) Experimental (b) Analytical
Figure 21. Failure pattern of double-leaf masonry wall on the front side of W7
Figure 22. Load-deflection curve for Wall 7
5 Conclusion
In this research seven tests have been performed, four on single-leaf walls, and three on double-
leaf walls (these being quite a rarity in existing literature), to evaluate the influence that a collar-jointed
masonry addition can provide to the deformation and strength performance of an isolated masonry
wall. Two variants for the retrofitting application were examined; either on a “fresh” single-leaf
undamaged wall (i.e. quoted as pre-earthquake) or on a lightly damaged single-leaf wall (i.e. quoted
as post-earthquake). Based on the results from the single-leaf walls, the mortar strength doesn’t have
a remarkable influence on the mechanical behaviour of masonry walls as different types of mortar
both result in a very similar failure pattern and failure load. Diagonal cracking was observed to be the
main failure pattern in both the single-leaf and double-leaf masonry walls under the application of an
external effectively diagonal loading. Compared to the single-leaf wall, the double-leaf masonry not
only has a higher capacity and ductility, but also improves the integrity (reducing crackingand residual
stiffness). For the scenario where the collar joint technique was applied as a pre-earthquake
enhancement method (discarding the differential aging parameter of the two masonry leafs) was
found very effective in terms of failure load and stiffness increase. When used as post-earthquake
retrofitting the technique was also quite effective particularly in terms of stiffness and ductility increase.
The apparent additional options of considering different degrees of initial damage, parametricaly
altering the mortar parameters and modifying the mechanical bond of the two leafs should be seen as
means of enriching this research which in this preliminary stage showed promising novel results.
Furthermore, a numerical micro-modelling approach has been carried out to validate the
experimental findings. The results although the bold character of some parameter idealisation showed
a relatively good agreement in aspects such as the ultimate load and stiffness while for most cases
under-predicted the capacity for substantial plastic deformation, a feature which is critical in any
earthquake related study where the hysteretic infill performance is critical in the holistic structural
performance.
Acknowledgements
The authors would like to thank the China Scholarship Council (CSC) and the University of
Leeds for providing the financial support of this project.
References
1. Mehrabi A, Shing BP, Schuller M, Noland J. Experimental evaluation of masonry-
infilled RC frames. Journal of Structural Engineering, ASCE 1996; 122(3):228-237.
2. Ehsani M, Saadatmanesh H, Velazquez-Dimas J. Behavior of retrofitted URM walls
under simulated earthquake loading. Journal of Composites for Construction, ASCE
1999; 3(3):134-142.
3. ElGawady MA, Lestuzzi P, Badoux M. A review of conventional seismic retrofitting
techniques for URM. In: Proc 13th international brick and block masonry conference.
2004.
4. ElGawady MA, Lestuzzi P. A review of retrofitting of unreinforced masonry walls
using composites. In: Proc 4th international conference on advanced composite
materials in bridges and structures. 2004.
5. ElGawady MA, Lestuzzi P, Badoux M. Retrofitting of masonry walls using shotcrete.
In: NZSEE Conference. 2006. No 45.
6. Bhattacharya S, Nayak S, Dutta SC. A critical review of retrofitting methods for
unreinforced masonry structures. International Journal of Disaster Risk Reduction
2014; 7:51-67.
7. Alcaino P, Santa-Maria H. Experimental response of externally retrofitted masonry
walls subjected to shear loading. Journal of Composites for Construction ASCE
2008; 12(5):489-498.
8. Mahmood H, Ingham JM. Diagonal compression testing of FRP-Retrofited
unreinforced clay brick masonry wallettes. Journal of Composites for Construction
ASCE 2011; 15(5):810-820.
9. Anand SC, Yalamanchili KK. Three-dimensional failure analysis of composite
maosnry walls. Journal of Structural Engineering ASCE 1996; 122(9):1031-1039.
10. Pina-Henriques J, Lourenco PB, Binda L, Anzani A. testing and modelling of
multiple-leaf maosnry walls under shear and compression. In: Modena C, Lourenco
PB, Roca P, editors. Proceedings of the 4th international seminar on structural
analysis of historical constructions, Balkema: Leiden (NL); 2004; 1: p.299-310.
11. Ramalho MA, Taliercio A, Anzani A, Binda L, Papa E. A numerical model for the
description of the nonlinear behaviour of multi-leaf masonry walls. Advances in
Engineering Software, 2008; 39(4):249-257.
12. Binda L, Pina-Henriques J, Anzani A, Fontana A, Lourenco PB. A contribution for the
understanding of load-transfer mechanisms in multi-leaf masonry walls: testing and
modelling. Engineering Structures 2006; 28(8):1132-1148.
13. British Standard Institution, BS 5628-1:2005. Code of practice for the use of
masonry-Par1: Structural use of unreinforced masonry.
14. Peraza D. Special problems with composite multiwythe masonry walls. Forensic
Engineering 2009; 66-73.
15. Lourenco PB. Computations of historical masonry constructions. Progress in
Structural Engineering and Materials 2002; 4(3):301-19.
16. Chaimoon K, Attard MM. Modeling of unreinforced masonry walls under shear and
compression. Engineering Structures 2007; 29(9):2056-2068.
17. Page AW. Finite-Element Model for Masonry. Journal of the Structural Division,
ASCE 1978; 104(8):1267-1285.
18. Lourenco PB. Computational strategies for masonry structures. PhD thesis.
Netherlands: Delft University of Technology; 1996.
19. Al-Chaar GL, Mehrabi A. 2008. Constitutive models for nonlinear finite element
analysis of masonry prisms and infill walls (No. ERDC/CERL TR-08-19), US Army
Corps of Engineering.
20. Van Zijl GPAG. Modelling masonry shear-compression: role of dilatancy highlighted.
Journal of Engineering Mechanics 2004 130:1289-1296.
21. Dolatshahi KM, Aref AJ. Two-dimensional computational framework of meso-scale
rigid and line interface elements for masonry structures. Engineering Structures
2011; 33(12):3657-3667.
22. Cundall PA. A computer model for simulating progressive large scale movements in
blocky rock systems. In: Proceedings of the symposium of the international society
of rock mechanics, Nancy, France; 1971.
23. Lemos JV. Discrete element modelling of masonry structures. International Journal
of Architectural Heritage 2007; 1:190-213.
24. Sarhosis, V, Sheng, Y. Identification of material parameters for low bond strength
masonry. Engineering Structures 2014; 60:100-110.
25. Sarhosis, V, Garity SW, Sheng Y. Influence of brick-mortar interface on the
mechanical behaviour of low bond strength masonry brickwork lintels. Engineering
Structures 2015; 88:1-11.
26. Midas FEA. Nonlinear FE Analysis Software. MIDAS Information Technology Co,
(2013) CSP FEA.
27. British Standard Institution, BS 3921:1985. Specification for Clay bricks.
28. British Standard Institution, BS EN 772-1:2011, Methods of test for masonry units,
Part 1: Determination of compressive strength.
29. British Standard Institution, BS EN 1015-11:1999, Methods of test for mortar for
masonry-Part 11: Determination of flexural and compressive strength of hardened
mortar.
30. Vermeltfoort AT, Raijmakers TMJ. Shear tests on masonry walls. In: Proc 6th North
Americal masonry conference. 1993; p.1183-1193.
31. Van der Pluijm R. Material properties of masonry and its components under tension
and shear. In: Proc 6th Canadian masonry symposium, Saskatchewan, Canada,
1992; p.675-686.
32. Ghaboussi J, Barbosa R. Three-dimensional discrete element method for granular
materials. International Journal for Numerical and Analytical Methods in
Geomechanics 1990 14:451-72.
33. Oliveira DVC. Experimental and numerical analysis of blocky masonry structures
under cyclic loading. PhD thesis. Portugal: University of Minho; 2003